Previsão de criação de valor corporativo em empresas do G20 a partir de informações contábeis: análise empregando algoritmos de inteligência artificial

Abstract

The study presented in this article aimed to predict the Market Value Added (MVA®) of companies from G20 countries based on artificial intelligence (AI) algorithms, based on accounting variables. To this end, data from 16,452 public companies from G20 member countries between 2019 and 2022 were collected through the Refinitiv® Eikon Platform. Forecast models were estimated based on multiple linear regression (RL) and two algorithms AI: random forest (RF) and artificial neural networks (ANN). The following techniques were used for data analysis: regression analysis, descriptive statistics and Mann-Whitney test. In general, it was found that accounting variables were relevant to explain corporate value even in periods of crisis (Covid-19 pandemic and Russia's invasion of Ukraine). All models were considered significant and with high explanatory power of MVA®. In some periods, ANN-based models presented statistically superior results to those based on RL. Furthermore, there were no significant differences between the models estimated for developed countries in relation to emerging countries. In view of the above, the research presented in this article contributes significantly to the literature on Accounting and Finance from different aspects: (a) the relevance of accounting variables for predicting corporate value was highlighted, even in periods of crisis; (b) the importance of using AI algorithms to analyze this phenomenon in studies in the area was highlighted; and (c) the results were corroborated consideringthousands of companies from the most industrialized countries in the world, both developed and emerging.